By Pamela Martinez
AI stocks are under pressure, but the bigger question is whether the industry’s extraordinary infrastructure spending can deliver returns quickly enough to justify its cost.
As Nvidia prepares to report earnings, investors are looking beyond chip demand to the economics of the entire AI boom. The artificial intelligence boom has been built on an extraordinary investment proposition: spend heavily today on chips, data centres and computing infrastructure, and capture the economic returns of a technology expected to reshape entire industries tomorrow.
The latest sell-off in technology shares has brought that question into sharper focus. In early November, investors wiped approximately US$2.5 trillion from the market value of the Nasdaq 100 in a sharp retreat from AI-linked stocks. Semiconductor shares also came under pressure, while lenders and credit investors began reassessing their exposure to companies financing the data-centre expansion.
Nvidia, the industry’s most influential chipmaker, has been caught in the turbulence. Its upcoming third-quarter fiscal 2027 results, scheduled for 17 November, will be closely watched as a measure of whether demand for AI computing remains strong enough to support the industry’s enormous investment cycle. But daily share-price movements tell only part of the story. The more consequential issue is whether the economics of AI infrastructure can sustain the expectations built into its valuation. The sell-off may be a temporary correction, but it is also exposing questions about capital expenditure, financing costs, regulation and the time required to turn AI adoption into durable profits.
The Market Is Questioning the Price of the AI Boom
According to Reuters’ reporting on the November sell-off, the Nasdaq 100 lost approximately US$2.5 trillion in market value as investors reconsidered the outlook for data-centre development following the US midterm elections. Concerns included the possibility of greater scrutiny of data-centre projects, financing pressures and uncertainty about the pace of future investment.
Political uncertainty was only one part of the picture. In a separate report, Reuters found that banks and credit investors were becoming more cautious about AI-related debt. Lenders were reassessing their exposure to data-centre operators and AI infrastructure companies, with some showing reluctance to finance additional deals even when major technology companies were involved. This matters because the AI investment cycle is no longer funded exclusively by the cash generated by established technology companies. It increasingly involves debt, equity issuance, long-term contractual commitments, leases and more complicated financing arrangements.
The International Monetary Fund’s October 2026 Global Financial Stability Report identifies an AI-driven market correction as a potential risk to global financial stability. The IMF does not conclude that the sector is necessarily in a bubble. It notes that leading AI companies have delivered strong earnings and that their underlying businesses remain profitable. However, weaker or less durable AI investment and profitability could trigger a correction, particularly as leverage and interconnected financial exposures increase.
The Infrastructure Bill Is Becoming Harder to Ignore
The scale of AI investment helps explain why the market has become more sensitive to financing and profitability. According to J.P. Morgan Global Research, global AI capital expenditure could reach approximately US$900 billion in 2026. The bank estimates that the five major US hyperscalers are on track to spend roughly as much on AI capital expenditure in 2026 alone as they have spent cumulatively since ChatGPT launched in November 2022.
The investment is not without evidence of demand. Cloud providers are reporting substantial revenue growth, and contracted cloud-service demand provides an indication of future business. J.P. Morgan identifies cloud revenue as one of the clearest early indicators that AI infrastructure spending is beginning to generate returns.
In its August 2026 analysis, S&P Global Ratings projected that combined hyperscaler capital expenditure would exceed US$1.3 trillion by 2027. It also expected the six largest hyperscalers covered in its analysis to generate negative free operating cash flow in 2026 and 2027, with recovery not projected until 2029. Debt, leases, guarantees and other financing structures are becoming increasingly important to the expansion.
Nvidia’s Earnings Will Test the Strength of AI Demand
Nvidia occupies a particularly important position in this debate because its performance provides a view into the spending decisions of the companies building AI infrastructure. These figures are difficult to reconcile with a simple narrative that AI demand has suddenly disappeared. Nvidia continues to benefit from the need for advanced computing capacity, while its customers are investing heavily in the infrastructure required to develop and operate AI systems.
Investors will also be looking for evidence that customers can turn their infrastructure spending into sustainable business returns. Are cloud providers generating enough revenue from AI services to justify the cost of building new capacity? Are enterprises moving beyond experimentation towards regular, revenue-generating use? Are customers willing to commit to long-term computing contracts, and can suppliers meet that demand without creating excess capacity?
Nvidia’s results cannot answer every question about the economics of AI. The company sells critical infrastructure, but the returns on that infrastructure ultimately depend on the businesses using it. A strong quarter for chip sales does not automatically prove that every data-centre project or AI application will be profitable.
There is also a company-specific complication. Recent scrutiny of Nvidia’s leadership has added another source of uncertainty to its share-price performance. The board’s investigation involving CEO Jensen Huang has coincided with the wider market sell-off, making it important not to attribute every decline in Nvidia shares to concerns about AI spending alone. The broader investment question remains whether demand for computing will translate into profitable, recurring revenue across the ecosystem.
The Next Phase of AI Will Be Judged by Returns, Not Spending
The IMF’s October 2026 analysis of the AI ecosystem warns that interconnected financing arrangements could amplify financial shocks if AI monetisation or productivity gains disappoint. It also identifies technological obsolescence and market concentration as vulnerabilities. Infrastructure built around today’s most advanced systems may not retain its value if computing requirements, model architectures or customer preferences change.
Regulation and public acceptance add further uncertainty. Data-centre projects depend on local planning decisions, grid access, energy prices and environmental requirements. The US midterm results have intensified debate about whether the rapid expansion of AI infrastructure should face stronger oversight. Even where demand remains strong, delays or higher compliance costs could change project economics.
For investors, the distinction between the strongest companies and the most exposed ones may become increasingly important. Large technology groups with established revenue streams and diversified businesses may be better positioned to absorb high capital expenditure. Smaller infrastructure providers and companies relying heavily on borrowing may be more vulnerable if financing becomes expensive or customer commitments weaken.
The sell-off should not be read as proof that AI has failed. Nor should continued growth in chip revenue be treated as proof that every investment in the sector will succeed. The more useful interpretation is that the market is moving into a more demanding phase of the AI cycle. The technology’s potential is no longer enough to justify every valuation, financing arrangement or infrastructure project. Companies will have to demonstrate that the economic value created by AI can keep pace with the cost of building it.
Nvidia’s upcoming earnings may provide an important signal. But the longer-term test will be whether businesses across the AI ecosystem can turn unprecedented investment into durable productivity, cash flow and profit. The next AI winners may not simply be the companies that spend the most. They will be the ones that can prove the spending was worth it.
